tsset countycode year, yearly **Tests for serial correlation within panels (Baseline Model - Model I in Table 2) xtserial diffaccess diffmedicaid hispanicproportion blackproportion americanindianincludedalaskannat other difftobaccouse diffmateducation diffpopdensity timetrend benton- yamhill **Tests for panel heteroskedasticity (Baseline Model - Model I in Table 2) xtgls diffaccess diffmedicaid hispanicproportion blackproportion americanindianincludedalaskannat other difftobaccouse diffmateducation diffpopdensity timetrend benton- yamhill, igls panels(heteroskedastic) estimates store hetero xtgls diffaccess diffmedicaid hispanicproportion blackproportion americanindianincludedalaskannat other difftobaccouse diffmateducation diffpopdensity timetrend benton- yamhill local df = e(N_g) - 1 lrtest hetero . , df(`df') **Models for the paper **Model I, Table 2 xtreg diffaccess diffmedicaid hispanicproportion blackproportion americanindianincludedalaskannat other difftobaccouse diffmateducation diffpopdensity timetrend, fe cluster(countycode) **Model II, Table 2 xtreg diffaccess diffmedicaid hispanicproportion blackproportion americanindianincludedalaskannat other difftobaccouse diffmateducation diffpopdensity timetrend diffmedicaidhispanic diffmedicaidafrican diffmedicaidNA diffmedicaidother, fe cluster(countycode) **Model III, Table 2 xtreg diffaccess diffmedicaid hispanicproportion blackproportion americanindianincludedalaskannat other difftobaccouse diffmateducation diffpopdensity diffpoverty timetrend, fe cluster(countycode) **Model IV, Table 2 xtreg diffaccess diffmedicaid hispanicproportion difftobaccouse diffmateducation diffpopdensity timetrend, fe cluster(countycode) **SUPPLEMENTARY MODELS (ROBUSTNESS CHECKS NOT SHOWN IN PAPER) **Model with self-financing xtreg diffaccess diffselfpay hispanicproportion blackproportion americanindianincludedalaskannat other difftobaccouse diffmateducation diffpopdensity timetrend, fe cluster(countycode) **Model with private insurance xtreg diffaccess diffprivateinsure hispanicproportion blackproportion americanindianincludedalaskannat other difftobaccouse diffmateducation diffpopdensity timetrend, fe cluster(countycode) **Model with changes in physicians per 1000 rather than population density as proxy of care availability xtreg diffaccess diffmedicaid hispanicproportion blackproportion americanindianincludedalaskannat other difftobaccouse diffmateducation diffphysicians timetrend, fe cluster(countycode)